Amazing energy at the @CFTC today! Great to see TradFi OGs, crypto survivors and new prediction market leaders together to discuss strengthening U.S. leadership in global financial markets and keeping innovation on shore versus driving it away. It is exciting to see the Administration meeting the moment -- thanks for your leadership @MichaelSelig and for your work in facilitating this conversation @WaltFIA. As I said yesterday, we are in a global race for AI supremacy. Compute futures will lower the cost of capital in the space, thereby extending our competitive advantage. @Silicon_Data has developed well-constructed indices that will serve as the underlying for new futures contracts listed on @CMEGroup. This is a generational opportunity and the sooner we launch them, the better for America.

Everybody saw it in the market. The @Silicon_Data indices all put in lows around then and since have been up and to the right. However not all GPU indices are equally well constructed. This graph highlights a serious discrepancy. I know which futures contract I will be using.

A big week for @Silicon_Data — and two strong signals of how quickly compute is growing in importance. Its partnership with @CMEGroup to bring compute futures to market demonstrates the demand for tools for price discovery and managing risk. And with its $30.5M Series A, Silicon Data is well positioned to deliver the infrastructure and tools needed to keep pace with this rapidly developing market. Huge credit to @carmenli and her team. From a macro perspective, investment in ‘compute refineries’ is far outpacing investment in traditional refineries, which is another sign that my forecast of more dollars per year spent on compute than oil is coming closer to fruition (and maybe even more quickly than my 10 year prediction!).

Survived my first CNBC hit at 3am HK time. Lineup: @cerebras CFO @BobKomin on IPO day, @OctahedronCap, @SambaNovaAI CEO… and me, talking the first compute futures market with @CMEGroup Thanks @dee_bosa for having me.

I've believed for years that compute would evolve into one of the world’s most important commodities — which is why I backed the creation of @Silicon_Data and @computeexchange two years ago. Today’s announcement from @Silicon_Data and @CMEGroup is an important step in that direction. As AI scales, compute markets are developing the same kinds of supply, volatility and capital allocation dynamics we’ve seen in energy and other major commodity markets. Futures markets matter because they improve price discovery, reduce the cost of capital and support long-term infrastructure investment.

Digital Asset Holdings, the company behind a blockchain used among several big banks and trading firms, is raising money from investors including a16z crypto at around a $2 billion valuation, people with knowledge of the matter said

Tokenised repo only works if the “cash” leg is truly cash-equivalent. Some thoughts on how proposed US stablecoin rules could impact that, and why transparency in reserves will be critical for market confidence. Thank you to @ISDA for hosting an important conversation on how proposed stablecoin rules could shape liquidity and market structure.

The bubble debate is a distraction. Compute already has natural longs, natural shorts, speculators, a forward curve, and benchmarks. What it doesn't have yet: standardized contracts and deep hedging venues. The connective tissue every other commodity figured out 50 years ago. That gap is what's actually pricing risk in the market right now. David Friedman, Dmytro Lokshyn and I are going deep on it next Thursday. Thursday, May 7, 2026 — 12:00 PM ET (9:00 AM PT)  Live online via Zoom Link:

I am excited about the launch of options on Eris swap futures on June 16! Congratulations @ErisFutures on the dramatic growth in volume and open interest of the futures, which has laid the groundwork for the launch of options. Thank you for the support from @CMEGroup!

No, aggregate properties of private assets, such as total supply cap, cannot be verified by every user on Canton yet. If this is the main criticism of Canton, I wholeheartedly accept it! All of the claims people are throwing on X about Canton not being decentralized, not being cryptographically verifiable, and requiring complete trust in institutions are red herrings. Your observation is the only correct critique of Canton I've seen yet. I agree that publicly proving aggregate amounts is a desired property and that all transparent chains have it. zkSync Prividiums also have it, under the assumption that the entire ZKP stack works as advertised. Canton does not (yet) have this property. Now that we've distilled it to the core property that Canton lacks, we can talk about why the importance of it is completely overblown, and why this hasn't been a concern in practice for any users: (1) Many privacy assets on Canton have decentralized BFT issuers. You're not trusting any single institution; you're trusting a BFT consensus. A single honest Validator can report and cryptographically prove dishonest behavior. (2) For asset-backed tokens, Canton has the same trust model as transparent public chains. You ALREADY rely completely on an independent auditor to review that the aggregate amount on-chain equals the off-chain amount. Those same auditors cryptographically and independently audit the on-chain amount in Canton. We're already trusting stablecoin providers + their auditors to maintain the on- vs off- chain reserves, so the trust model on Canton is exactly the same! But Canton makes it much better - we're moving to a world where tokenized US Treasuries are on-chain, so the auditor gets cryptographic independent verification of the peg. If anything, Canton is the ONLY blockchain bringing on-chain cryptographic verifiability of RWA reserves! Repo transactions on Canton are the only repo transactions in the world where there's distributed cryptographic verification that your asset is fully backed! (3) For fully dematerialized assets, the registrar is legally allowed to change the aggregate amount. The important thing is that it's auditable by the issuer, not that it's publicly auditable. If DTCC tells you that Tesla has 3.3b shares outstanding, even if you could verify that aggregate number on-chain, can you independently verify that that's the correct amount? Only Tesla can say whether that number is correct so they're the only voice that matters from a trust-model perspective. All that said, I agree that publicly proving aggregate properties is a nice-to-have. But it's never been desirable enough to justify the trade-off of adding significant complexity to the software stack. Running a Canton Validator in a TEE would be dead simple and give you that property (with different trust assumptions), yet in practice, no one seems to think it's important enough to justify even that modest investment. This is actually one area where retrofitting ZKPs on top of Canton would not be very hard! By relaxing this single requirement, Canton has been able to add many other security properties.

The only honest people about absolute permissionlessness are Bitcoiners. They wanted to solve a specific problem and stuck it through the test of time. If you’re into RWAs (we are and always have been and are unashamed of it) and claim to be all about absolute permissionlessness, you’re just being dishonest to pump a bag and nothing else. Once you want centralized issuers to operate in your ecosystem in a way that works for their business, you start to compromise. Enjoy the coffee.

In the 1980s, banks lent against commercial real estate without checking for contamination. Then the losses came. Then Phase I environmental assessments became standard overnight. David Friedman's latest piece makes the case that GPU-backed lending is at that same inflection point — and he builds the argument on research Yuhua Yu and I did at @Silicon_Data. The finding that stuck with me most: individual chip identity explains 32–73% of performance variance. Not the provider. Not the config. The chip itself. Two identical #H100s# can differ by up to 38% in real throughput. If you're structuring debt against GPU pools, the spec sheet isn't your collateral. What that hardware actually does under load is. Great working with Dave on this one.